Key result
RapidAI yields lower overall ASPECTS agreement than neuroradiologists, but improves agreement in extensive ischemia.
Why the study?
Despite widespread adoption of RapidAI for acute stroke management, gaps remain in validating its real-world diagnostic impact, compounded by local observations of RapidAI overestimating infarct size for equivalent images.
Does RapidAI improve agreement with follow-up imaging for Alberta Stroke Program Early CT score (ASPECTS) compared to neuroradiologist interpretation in patients with acute ischemic stroke?
Cohort (n=481)
No
Does RapidAI improve agreement with follow-up imaging for Alberta Stroke Program Early CT score (ASPECTS) compared to neuroradiologist interpretation in patients with acute ischemic stroke?
Neuroradiologists maintain higher overall agreement with follow-up imaging than RapidAI in acute MCA ischemic stroke, particularly in mid-range ASPECTS and deep infarcts, indicating RapidAI should be used cautiously as an adjunct rather than a replacement for expert interpretation.
Answers are grounded in this paper and cite their sources.
RapidAI ASPECTS shows no significant bias versus follow-up; leaves open routine adoption in stroke triage pending prospective validation.
Background Despite RapidAI's widescale adoption for acute stroke management, gaps remain in validating its real-world diagnostic impact. Given local observations of RapidAI overestimating infarct size for equivalent images, we evaluated the reliability of RapidAI versus neuroradiologists at a Canadian comprehensive stroke centre. Methods CT head and carotid studies (Jan 2020–Aug 2021) from acute stroke activations were retrospectively reviewed. Repeat presentations, no follow-up imaging within 10 days, bilateral infarcts, non-MCA strokes, or hemorrhage were excluded. Alberta Stroke Program Early CT score (ASPECTS) from RapidAI and neuroradiologists were compared to follow-up MR or CT imaging using intraclass correlation coefficient (ICC). Linear mixed-effects model was adjusted for age and sex. Subgroup analyses examined infarct burden and infarct location (cortical vs. deep). Results 481 patients (49% female; median age 74 years) were included. Neuroradiologists had higher agreement with the reference standard compared to RapidAI (ICC: 0.627 vs. 0.546). RapidAI had a smaller adjusted difference (β=-0.10; p=0.198) compared to neuroradiologists (β=0.74; p < 0.001). Although agreement was poor across both groups, RapidAI showed higher agreement in extensive ischemia (ASPECTS 0–4: 0.194 vs. 0.115; ASPECTS 0–2: 0.083 vs. 0.059), whereas neuroradiologists had higher agreement in threshold ASPECTS 5–7 cases (0.141 vs. 0.068). Neuroradiologists showed superior agreement in deep infarcts (ICC: 0.612 vs. 0.452). Conclusion Neuroradiologists maintain higher overall agreement in mid-range ASPECTS cases and deep infarcts, but RapidAI exhibited higher agreement in lower ASPECTS cases. RapidAI demonstrates context-dependent accuracy, serving as an adjunct requiring cautious use alongside neuroradiologist interpretation for clinical judgement.
No takes yet. Share an insight, caveat, or question.
Phung et al. (2026) conducted a cohort in Acute ischaemic stroke (n=481). RapidAI ASPECTS interpretation vs. RapidAI-assisted neuroradiologist ASPECTS interpretation was evaluated on Agreement of Alberta Stroke Program Early CT score (ASPECTS) with follow-up MR or CT reference standard using intraclass correlation coefficient (ICC) (ICC 0.546 vs 0.627). RapidAI showed lower overall ASPECTS agreement with follow-up imaging than neuroradiologists (ICC 0.546 vs 0.627), but higher agreement in extensive ischemia (ASPECTS 0-4: 0.194 vs 0.115).
Synapse has enriched 10 related papers. Consider them for comparative context: